Compliance & trust
Security questionnaires, control evidence, model and agent audits: each answer tied to its source, with a signed record a reviewer can re-check. Open a job to see a real run first, then run it yourself on a sample.
Make and do
Check and prove
- Map evidence to NIST 800-171A gap map of all 320 NIST 800-171 objectives, each marked present, partial or missing with quotes from your evidence, plus a draft POA&M.237 / 286Objectives given the right status: present, partial or missing (held-out test)Try it
- Check your provider serves the model you pay forPaying a provider for “Qwen 27B at BF16”?Find out whether an OpenAI-style endpoint really serves the model it claims, or a smaller or more compressed one. You get a signed pass, drift or fail report you can put in the vendor file, and anyone can re-check it.~$0.18 per 100 auditsmeasured, at list priceTry it
- Prove what your agent didCompliance asks what your agent did, and why.A signed, step-by-step record of what your agent saw, decided and did, which anyone can re-check. Change one step later and verification names it.339/339Edited copies of a record caught, signing key pinned (synthetic set)Try it
- Check green claims in copyA table of every environmental claim in your copy: banned, needs evidence or backed, with the rule, the evidence line and a safer rewrite.43/44 (98%)Claims given the right verdict (held-out set, first run)Try it
- Draft VEX for scanner findingsA draft VEX statement for every scanner finding, with the evidence from the image itself, ready for a security engineer to sign off.154 / 159'Not affected' calls that match Canonical's own statements (held-out test)Try it
More coming
We add jobs here as each one is built and checked on real runs.
More tools
- Catch AI answers your sources don't backEvery sentence of an AI-drafted answer checked against the source passage it relies on, with the quote, and a pass, flag or block you can put in front of Send.Try it
- Draft a tariff classification memoA classification memo with a proposed HTS code, the rules applied and word-for-word quotes from similar CBP rulings, or a clear 'needs a broker'.Try it
- Find accessibility fixes for a shopA ranked fix list of accessibility problems that stop shoppers buying, each with its WCAG criterion, the element and a code fix.Try it
- Check if a video is made for kidsA factor-by-factor answer on whether an upload is aimed at children, with the lines that show it, whether your setting matches, and what to change.Try it
- Build a model-risk evidence packLabsAn evidence pack showing whether an LLM you rely on still behaves as validated, from which your validators can recompute every number.Try it
- Draft breach notices and deadlinesLabsA timeline, a deadline per regime (8-K, NIS2, DORA and more), and each notice checked sentence by sentence against the incident log.Try it
- Make a tamper-evident meeting recordLabsLive captions, then a speaker-labelled transcript and a summary whose every sentence cites transcript lines, in a signed record that shows any edit.Try it
- Log human edits and editor sign-offLabsA signed record per piece showing the AI draft, every human edit and the named editor's sign-off, with a public page readers can check.Try it
- Pre-check promo claims for MLRAn annotated MLR pre-review packet: each claim with the reference passage behind it, plus off-label, comparative, fair-balance and disease-claim problems.Try it
- Run an end-to-end browser testA browser run of your plain-language test, judged only by your assertions in code, ending in a signed certificate that fails the build on a failure.Try it
- Check an AI noteEvery note sentence checked against the transcript, with the line it came from or 'not in the transcript': invented facts, changed doses and details, exam findings nobody made, the wrong speaker, and key items left out. Plus a shorter note that only removes words, and a signed scorecard per vendor.Try it
- Keep a signed lab notebookLabsA tamper-evident notebook that records which conclusions an AI model helped with, from which data files, and that an inspector can verify in the browser.Try it
- Write a SAR narrativeA SAR narrative in FinCEN's structure where every sentence cites its rows and every amount, date and count is recomputed from them.Try it
- Review a claim file for conductA signed review of each claim file: every deadline worked out from the file's own dates, and each denial reason checked against the policy wording.Try it
- Review a collections callA yes, no or unclear answer to each Regulation F and X call question, with the quote and its time, plus call-log breaches computed from your log.Try it
- Tie out MD&A figuresLabsA workpaper that ties each MD&A figure to the table cell or calculation it came from, and flags the ones that do not match.Try it
- Disposition a sanctions alertA field-by-field comparison of the customer and the list entry, a proposed call with its rule, and a signed decision record naming the list version.Try it
- Triage a device complaint for MDRA triage memo answering the 803.50(a) questions with quotes from the complaint, a suggested call, the MDR deadline worked out, and a draft 3500A narrative.Try it
- Investigate a Reg E disputeAn investigation file per dispute: every Reg E deadline worked out with its rule, the notice's required details quoted, and the results letter checked.Try it
- Build an EU GPSR listingLabsThe EU listing fields pulled from your product sheet and label, each Article 19 item marked found or missing, and warnings translated with every number checked.Try it
- Check generated product imagesLabsA pass or stop for each AI product image, naming any wrong size, colour, text, logo, warning or extra part, and the AI label and metadata for images that pass.Try it
- Draft an adverse-event caseA draft E2B-shaped case with every field quoted to its source, seriousness and expectedness per event, and the 15- and 90-day reporting dates worked out.Try it
- Compare safety sections across labelsA list of every difference in the safety sections across the label documents and translations, quoted from each and marked likely error or deliberate.Try it
- Answer a payer auditThe respond-by date, each claim checked against every documentation requirement of the payer's policy with the chart, page and line, the weak claims first, a draft cover letter and an indexed response packet.Try it
- Capture audit evidence from your admin screensPreviewEach named screen captured with URL, time and user stamped, every setting checked against the baseline you approved, changes first, in a signed pack.Try it
- Check a bank-detail change before you payPreviewThe warning signs with evidence, the call-back to the number already on file, and a signed record with a second approver.Try it
Looking for something else? Every tool, as an index.